Global Explainability of GNNs via Logic Combination of Learned Concepts